Elon Musk's SpaceXAI Launches Grok 4.5: Try It Free in Cursor Now

The newly rebranded SpaceXAI—which is what happens when you merge a rocket empire with a social media billionaire's AI side-hustle—just unleashed Grok 4.5, built on a fresh 1.5-trillion-parameter base model called V9. Instead of training it in a vacuum, they literally bought their way into the developer ecosystem by acquiring Anysphere, the creators of the Cursor code editor, for a mind-boggling sixty billion dollars.

This absurdly expensive shopping spree allowed SpaceXAI to inject Cursor's actual user data straight into the AI's veins, with the editor's engineering team actively helping to train the model. Developers can now test drive the model for free inside Cursor and Grok Build, while the API will set back power users a modest two dollars per million input tokens and six bucks for output.

In a shocking twist of corporate honesty, the marketing team didn't even try to fake the benchmarks to make themselves look like undisputed gods. Grok 4.5 failed to take the absolute crown, losing three out of five major tests to Anthropic's heavyweight champion, Fable (max).

While it managed to squeak past Opus 4.8 (max) on terminal and basic code tasks, it fell flat on multilingual and advanced software engineering benchmarks. However, the real flex here is economics: Grok 4.5 screams at eighty tokens per second and solves complex coding problems using over four times fewer output tokens than its competitors.

Buying an entire developer ecosystem just to prove an AI can write code cheaper than the competition is peak tech billionaire behavior. It seems the future of software engineering isn't about who writes the smartest model, but who has the deepest pockets to subsidize the API fees until everyone else goes broke.

Source: xAI